Desenvolvimento de um sistema de controle para um digitalizador de armadilhas adesivas

Abstract

Aphids are insects considered pests in agriculture. The management and control of these agricultural pests are essential to ensure the vitality of crops and better crop yields. Computer vision projects aim to automate this process. However, these programs require high-resolution images acquired by scanners. This paper presents a control system for a sticky trap scanner that aims to automate the image acquisition process, sending the acquired images to the BIS insect monitoring platform where they are automatically processed by the intelligent aphid detection and counting model InsectCV. The system also has the option of saving the images in cloud storage services to be used in training new detection models. A proof-of-concept version of the scanner was designed and produced for later application for an invention patent. Based on the tests performed, it was possible to validate that the images generated by the system have the necessary quality to be used by intelligent models. However, adjustments are still needed in the focus control and illumination at the edges of the acquired images, to prevent loss of parts of the image.